Hi All,
Can anyone tell me why VMware recommends fixed policy for Active active storage and MRU for active passive?
Check if this kb answer your question http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=101134...
Please, don't forget the awarding points for "helpful" and/or "correct" answers.
Mauro Bonder - Moderator
Check if this kb answer your question http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=101134...
Please, don't forget the awarding points for "helpful" and/or "correct" answers.
Mauro Bonder - Moderator
The KB article doesnot mention " why". It explains what it is. Could you please tell me why I have to use MRU for A/P array and fixed for A/A array
http://www.vmware.com/pdf/vsphere4/r40/vsp_40_san_cfg.pdf
when there is a fiber board by accessing a storage controller, depending on the storage when a failover is just changing the storage controller where there will be access the LUN.
The Fixed can also be used to make manual balancing through the storage, requesting a lun for each storage.
Please, don't forget the awarding points for "helpful" and/or "correct" answers.
Mauro Bonder - Moderator
Why it's simple...
For a active/standby contollers the LUN migration is an expensive function (usually its is called tresspass and may be notified by the storage) and require TIME... this mean that I/O operations are delayed for some seconds (time depends by the controller type).
To avoid this problem this operation must be used only in case of failure.
Fixed option, instead define a fixed path (possible different for each LUN or at least balanced) and this path must be used for performance aspect.
Andre
I have similar question about Fixed policy for A/A array. If using Fixed policy for A/A array, only one path is active for I/O, I don't think it fully utilizes A/A array's great performance, since for A/A array, all ports on two controllers can go through I/O accessing one LUN in the array. Fixed policy doesn't provide load balancing actually.
So why not RR policy is the better choice for A/A array? Does RR policy really realize load balancing for A/A array? or so far, we can only actually direct I/O through one path because we are still using SCSI-2 protocal in our storage multipathing feature?
Looking forward to your reply.
Thanks,
Shannon